For ecommerce businesses, product discovery has a direct impact on conversion, basket size, and customer experience. Static product grids and manually curated recommendations can work at a small scale, but they often struggle to adapt to changing customer behaviour, catalogue complexity, and individual purchase intent.
This project involved implementing a personalised recommendation engine using Vertex AI for ecommerce. The goal was to use behavioural and catalogue data to present more relevant product suggestions across the customer journey. Data preparation was a critical first step. Product catalogue fields, event tracking, user interactions, purchase history, and product metadata needed to be structured so the recommendation system could learn from meaningful signals.
Recommendation placements were designed for high-impact ecommerce touchpoints, including the homepage, product detail pages, category pages, cart, checkout support areas, and post-purchase experiences. Each placement had a different role. Some were intended to increase discovery, while others supported cross-sell, upsell, replenishment, or alternative product exploration.
The implementation connected Vertex AI recommendations into the website experience and included a measurement framework to compare performance against existing recommendation logic. Metrics such as click-through rate, add-to-cart rate, product coverage, revenue per session, and average order value were monitored to understand commercial impact.
The final system helped customers find more relevant products with less effort. It also gave the ecommerce team a scalable personalisation capability that could continue improving as more behavioural data became available.